TFRServer Deployment

Deploy TFRServer service instances.

Overview

TFRServer is a namespace-scoped resource deployed in the tenant namespace created by TFRTenant, managing individual TFRobot server instances:

• API Server - Main service process
• Worker - Async task processing
• RobotWorker - Robot task processing
• Confz Config - Maps 1:1 to TFRobotServer TOML config; user-specified values override system-generated values
• Secret / ConfigMap - Third-party API Keys and environment variables

Prerequisites

TFRServer depends on the tenant namespace and middleware created by TFRTenant. Ensure TFRTenant status is Available before deployment.

---

Configuration

Core Fields

Field	Description	Example
name	TFRServer resource name (i.e. robotId)	friday
namespace	Tenant namespace (created by TFRTenant)	staging-tenant
confzData	Confz structured config	See below
secretData	Sensitive env vars (API Keys, etc.)	See below
configData	Non-sensitive env vars	See below

Confz Config

confzData maps 1:1 to TFRobotServer's TOML config. Most fields are dynamically generated by the system (postgresql, redis, celery, cos, grafana, etc.). Typically only the following fields need configuration:

confzData:
  admin:
    adminSecret: "<admin-secret>"
    jwtSecret: "<jwt-signing-secret>"
    password: "<argon2-hashed-password>"
  robot:
    robotId: "friday"
    host: "https://staging.turingfocus.cn"
    dbSchema: "tfrobot"
    sessionSecret: "<session-secret>"
  loguru:
    level: "INFO"
  feature:
    uploadFileSizeLimitMb: "50"

System-Generated Fields

postgresql, redis, celery, cos, grafana, otlp configs are auto-generated by the Controller based on TFRTenant middleware status. To override, explicitly specify them in confzData.

Image Configuration (Optional)

apiImage: "docker.cnb.cool/turingfocus/tfrobotv2/tfrobotserver:latest"
workerImage: "docker.cnb.cool/turingfocus/tfrobotv2/tfrobotserver:latest"
robotWorkerImage: "docker.cnb.cool/turingfocus/tfrobotv2/tfrobotserver:latest"
imagePullPolicy: "IfNotPresent"
imagePullSecret: "tfrs-private-registry"

Uses Controller's built-in default images when not specified.

Storage Configuration (Optional)

storage:
  storageClass: "cbs"
  defaultSize: "10Gi"

Inherits from TFRTenant or TFRCluster config when not specified.

---

Sample Configs

• config/samples/tfrobot_v1_tfrserver_staging.yaml - Staging environment (with full field comments)
• config/samples/tfrobot_v1_tfrserver.yaml - General template (TODO placeholders)

---

Kustomize Deployment

Create CR Manifest

# tfrserver-staging.yaml
apiVersion: tfrobot.tfrobotserver.cn/v1
kind: TFRServer
metadata:
  name: friday
  namespace: "staging-tenant"
spec:
  confzData:
    admin:
      adminSecret: "<admin-secret>"
      jwtSecret: "<jwt-signing-secret>"
      password: "<argon2-hashed-password>"
    robot:
      robotId: "friday"
      host: "https://staging.turingfocus.cn"
      dbSchema: "tfrobot"
      sessionSecret: "<session-secret>"
    loguru:
      level: "INFO"
    feature:
      uploadFileSizeLimitMb: "50"

  imagePullSecret: "tfrs-private-registry"

  secretData:
    OPENAI_API_KEY: "<your-openai-key>"
    ANTHROPIC_API_KEY: "<your-anthropic-key>"
    # ... other API Keys

  configData:
    TFS_PROFILE: "tkeprod"
    OPENAI_BASE_URL: "https://api.openai-proxy.org/v1"
    ANTHROPIC_BASE_URL: "https://api.openai-proxy.org/anthropic"
    TOKENIZERS_PARALLELISM: "false"

Full Field Reference

Full staging config example: config/samples/tfrobot_v1_tfrserver_staging.yaml

Deploy

kubectl apply -f tfrserver-staging.yaml
# tfrserver.tfrobot.tfrobotserver.cn/friday created

Verify

# Check TFRServer status
kubectl get tfrserver -n staging-tenant

# Wait for deployment to complete (timeout 10 minutes)
kubectl wait tfrserver friday -n staging-tenant --for=condition=Available --timeout=600s

# Check component status
kubectl get tfrserver friday -n staging-tenant -o jsonpath='{.status.components}' | jq

# Check service URL
kubectl get tfrserver friday -n staging-tenant -o jsonpath='{.status.serviceURL}'

Monitor deployment progress in real-time

watch -n 5 'echo "=== TFRServer Status ===" && \
  kubectl get tfrserver friday -n staging-tenant -o jsonpath="{.status.conditions[0].message}" && \
  echo && echo && \
  echo "=== Pods ===" && \
  kubectl get pods -n staging-tenant -l app.kubernetes.io/instance=friday'

Check Deployment Completion

TFRServer uses the Available condition in status.conditions to indicate deployment status:

# Check Available condition
kubectl get tfrserver friday -n staging-tenant -o jsonpath='{.status.conditions[?(@.type=="Available")].status}'
# True = deployment complete, False = deploying or failed

# Automation script: wait for deployment (timeout 10 minutes)
kubectl wait tfrserver friday -n staging-tenant --for=condition=Available --timeout=600s

Uninstall

kubectl delete tfrserver friday -n staging-tenant

---

Helm Deployment

Deploy

Install from OCI registry

helm install tfrserver oci://helm.cnb.cool/turingfocus/tfrs-operator/tfrserver \
  --version 0.2.0-alpha1 \
  --namespace staging-tenant \
  -f values-staging.yaml

Install with values file

helm install tfrserver oci://helm.cnb.cool/turingfocus/tfrs-operator/tfrserver \
  --version 0.2.0-alpha1 \
  --namespace staging-tenant \
  -f values-staging.yaml \
  --set confzData.admin.adminSecret="<SECRET>" \
  --set confzData.admin.jwtSecret="<SECRET>"

Install from local source

helm install tfrserver ./chart/tfrserver \
  --namespace staging-tenant \
  -f chart/tfrserver/values-staging.yaml

Note

TFRServer requires --namespace, which must match the namespace field in the values file — the tenant namespace created by TFRTenant.

Verify

# Check TFRServer status
kubectl get tfrserver -n staging-tenant

# Wait for deployment to complete (timeout 10 minutes)
kubectl wait tfrserver friday -n staging-tenant --for=condition=Available --timeout=600s

# Check component status
kubectl get tfrserver friday -n staging-tenant -o jsonpath='{.status.components}' | jq

# Check service URL
kubectl get tfrserver friday -n staging-tenant -o jsonpath='{.status.serviceURL}'

Monitor deployment progress in real-time

while true; do clear; echo "--- $(date +%H:%M:%S) ---"; \
  echo "=== Conditions ==="; \
  kubectl get tfrserver friday -n staging-tenant -o json 2>/dev/null \
    | jq -r '.status.conditions[]? | "\(.type): \(.status) - \(.message)"'; \
  echo; echo "=== Components ==="; \
  kubectl get tfrserver friday -n staging-tenant -o json 2>/dev/null \
    | jq -r '.status.components // {} | to_entries[] | "\(.key): \(.value.state)"'; \
  echo; echo "=== Pods ==="; \
  kubectl get pods -n staging-tenant -l app.kubernetes.io/instance=friday 2>/dev/null; \
  sleep 10; done

Press Ctrl+C to exit.

Check Deployment Completion

TFRServer uses the Available condition in status.conditions to indicate deployment status:

# Check Available condition
kubectl get tfrserver friday -n staging-tenant -o jsonpath='{.status.conditions[?(@.type=="Available")].status}'
# True = deployment complete, False = deploying or failed

# Automation script: wait for deployment (timeout 10 minutes)
kubectl wait tfrserver friday -n staging-tenant --for=condition=Available --timeout=600s

Uninstall

helm uninstall tfrserver -n staging-tenant

Note

Uninstalling TFRServer removes the service instance and its associated Secrets, ConfigMaps, Jobs, etc. Database data is preserved in TFRTenant's PostgreSQL.